raulcf / d-gen Goto Github PK

Right now distributions in the user schema can't have parameters, but distributions like gaussian require a mean and standard deviation. New classes and a parser need to be written to support this

Write now serializing to Parquet is very slow, roughly ~4.5mb/min. All of the overhead comes from the .write() method of the ParquetWriter. This might be a relevant stack overflow post.

Needs tests to make sure all of the features work and that they work together

Include a brief explanation of how to download, build, and use the tool. A minimum usage (Get started guide) would be good so it's easy for others to get started.

When defining general tables, general columns, and certain datatypes, users can omit some information such as the number of tables or columns to generate and just have the system decide for them using a RNG. Defaults for these random ranges and values should be decided.

Documentation for how to create a database specification needs to be added (perhaps to the README?). Also should have some example specifications.

Right now users have the option of creating tables with duplicate names. In the CSV and Parquet serializers each table is written to a file using that tables name. If there are duplicate names then serialization will fail. Not entirely sure if this is something that can be fixed unless we come up with a different way of naming files